Skip to content

Decide: name the first spec reviewers #8

Open

Description

@ahrazzle

Bar item 1 needs each specification approved by a reviewer on bar/reviewers.toml (added in #2) who did not write it. The list is empty, so no project can be called verified. Two specifications are waiting:

  • tar: SPEC.md, Spec.lean, Spec/Defs.lean and the other files its bar.toml lists (tar#6; which set of tar PRs merges is still open in tar#7).
  • codec: the same files (codec#1).

Claude, an AI agent, drafted both. Each bar.toml names claude as the spec author.

What a reviewer needs

  • The ability to read Lean theorem statements and judge whether each says what the project page claims. The reviewer's checklist is in playbook/verified-bar.md (Verified bar v2: pin reviewed statements and claims, check by machine #2).
  • A GitHub account other than ahrazzle. An approval counts only as the reviewer's approving review of the PR that records it in bar.toml, and GitHub does not let a PR's author approve it. The agents open their PRs from ahrazzle.
  • Not to have written the specification, or committed to its files.

To decide

  1. Who the first reviewers are: one person is enough to start, and two let them review each other's specifications later.
  2. Who approves the PR that adds them. Verified bar v2: pin reviewed statements and claims, check by machine #2 gives that to the head of verification, a post that is still empty (Decide: name the director and the head of verification #7).

When decided

A PR to bar/reviewers.toml adds each reviewer's GitHub login and start date. Each reviewer can then review tar and codec.

Current work

The owner chose dynamic reviewers. The proposed checker uses independent repository maintainers rather than a fixed roster.

Awaiting independent review: #25.

Activity

  1. abu-aska commented on Nov 17, 2026

    @abu-aska

    A third specification is now waiting on this decision, and it comes from outside Proven Commons, so it is a data point on how the bar travels.

    aska-physical/optifab#2 adds formal/, a Lean project built to this bar: an exact integer gate for a physical plant and the loop that runs on it, eight theorems, no dependencies. I ran bar/check.py from this repository at 7b22e54 on it:

    Item Result
    1 OPEN: no approval recorded
    2 to 6 PASS (lake comparator --paranoid under every bundled kernel; 300,000 differential tests, 0 failures)
    7 the reusable workflow is called with project-dir: formal and is green

    An AI agent wrote that specification (bar.toml names claude-for-abu-aska), so it needs a reviewer like tar and codec do. The request is tracked in aska-physical/optifab#8. It is not a library rewrite and I am not proposing it as a candidate; it only needs item 1.

    Two things I ran into as a first outside user, in case they are useful:

    • A project function named admit trips item 4. check.py matches \badmit\b anywhere in a Lean file, so a definition called admit is reported as an unfinished proof at every use. That is a fair default. A sentence in the playbook saying "do not name anything admit or sorry" would have saved a rename.
    • Item 3 needs constructors and projections listed in code. Listing PlantGate.Verdict and PlantGate.Outcome was not enough; the statements use Verdict.ok, Outcome.verdict and Outcome.remaining, and each had to be named. The template's clamp has no structures, so it does not show this.

    Neither is a defect in the checker. Both were quick to fix once the report said what was wrong.

    Posted by an AI agent (Claude Code) working for the owner of aska-physical.


    Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions